CN110706804B - Application method of mixed expert system in lung adenocarcinoma classification - Google Patents
Application method of mixed expert system in lung adenocarcinoma classification Download PDFInfo
- Publication number
- CN110706804B CN110706804B CN201910782827.5A CN201910782827A CN110706804B CN 110706804 B CN110706804 B CN 110706804B CN 201910782827 A CN201910782827 A CN 201910782827A CN 110706804 B CN110706804 B CN 110706804B
- Authority
- CN
- China
- Prior art keywords
- model
- parameters
- hme
- algorithm
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 20
- 208000010507 Adenocarcinoma of Lung Diseases 0.000 title claims abstract description 14
- 201000005249 lung adenocarcinoma Diseases 0.000 title claims abstract description 14
- 238000012360 testing method Methods 0.000 claims abstract description 18
- 238000012549 training Methods 0.000 claims abstract description 15
- 238000012216 screening Methods 0.000 claims abstract description 7
- 238000002790 cross-validation Methods 0.000 claims abstract description 6
- 238000005457 optimization Methods 0.000 claims description 8
- FGUUSXIOTUKUDN-IBGZPJMESA-N C1(=CC=CC=C1)N1C2=C(NC([C@H](C1)NC=1OC(=NN=1)C1=CC=CC=C1)=O)C=CC=C2 Chemical compound C1(=CC=CC=C1)N1C2=C(NC([C@H](C1)NC=1OC(=NN=1)C1=CC=CC=C1)=O)C=CC=C2 FGUUSXIOTUKUDN-IBGZPJMESA-N 0.000 claims description 6
- 238000012795 verification Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000004069 differentiation Effects 0.000 claims description 3
- 206010058467 Lung neoplasm malignant Diseases 0.000 abstract description 7
- 201000005202 lung cancer Diseases 0.000 abstract description 7
- 208000020816 lung neoplasm Diseases 0.000 abstract description 7
- 230000006870 function Effects 0.000 description 27
- 238000002591 computed tomography Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 208000009956 adenocarcinoma Diseases 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 238000013179 statistical model Methods 0.000 description 3
- 239000013598 vector Substances 0.000 description 3
- 238000007476 Maximum Likelihood Methods 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000003745 diagnosis Methods 0.000 description 2
- 210000004072 lung Anatomy 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 208000009458 Carcinoma in Situ Diseases 0.000 description 1
- 206010028980 Neoplasm Diseases 0.000 description 1
- 201000011510 cancer Diseases 0.000 description 1
- 230000008602 contraction Effects 0.000 description 1
- 201000004933 in situ carcinoma Diseases 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Public Health (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Medical Informatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Primary Health Care (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Epidemiology (AREA)
- Pathology (AREA)
- Image Analysis (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
The invention provides an application method of a hybrid expert system in lung adenocarcinoma classification, which comprises the following steps: step 1: sorting the original data; step 2: screening characteristic data with classification significance by adopting LASSO algorithm; step 3: dividing the characteristic data into a training set and a testing set according to the proportion of 7:3; step 4: performing five-fold cross validation on an ME or HME system in a training set to obtain an optimal model of the ME or HME system; step 5: and testing the optimal model of the ME or HME system in a test set. To provide a hybrid expert system with better predictive performance for the type of lung cancer.
Description
Technical Field
The invention relates to the technical field of computer medical treatment, in particular to an application method of a hybrid expert system in lung adenocarcinoma classification.
Background
Lung cancer is one of the malignant tumors with highest morbidity and mortality in the world, and early accurate diagnosis of the lung cancer can save medical resources and greatly relieve pain of patients. Among the major histological types of lung cancer, adenocarcinoma is increasingly the leading cause of death. However, from Computed Tomography (CT) images, it is difficult for radiologists to distinguish the three major subtypes, invasive Adenocarcinoma (IAC), carcinoma in situ (AIS), and micro-invasive adenocarcinoma (MIA).
Some single statistical models and machine learning methods attempt to classify lung adenocarcinoma subtypes, such as Support Vector Machines (SVMs) and Convolutional Neural Networks (CNNs), but do not achieve good results. In order to improve the prediction precision of a single model, the invention utilizes a minimum absolute shrinkage and selection algorithm (LASSO) to screen features based on CT images and clinical features of a patient of lung adenocarcinoma, trains a hybrid expert system (ME) and a hierarchical hybrid expert system (HME) by using a small batch random gradient descent (mini-batch-SGD) and an RMSProp algorithm, and adds a regular term to the mixture to assist diagnosis of lung adenocarcinoma.
LASSO was first proposed by Robert Tibshirani in 1996, which obtained a more refined model by constructing a penalty function such that it compressed some regression coefficients, i.e. the sum of absolute values of the forcing coefficients was less than a certain fixed value; while some regression coefficients are set to zero. The advantage of subset contraction is thus retained, being a biased estimate of the processing of data with complex co-linearity.
Hybrid expert systems use divide-and-conquer ideas to combine a number of simple expert models to solve complex problems, where each expert model is given a greater weight in solving its own adept problem. At the earliest by
Michael I Jordan et al, 1991. Based on the mixed expert system, the layered mixed expert system constructs a multi-stage expert system. Initially, both systems were trained by maximum likelihood and gradient ascent methods. Until 1993, these two systems did not begin to solve using the Expected Maximum (EM) algorithm, greatly improving the convergence rate of the entire model.
The expectation maximization algorithm (EM) is summarized by Dempster et al and is an iterative algorithm for solving maximum likelihood estimates of probability model parameters containing hidden variables. The method can estimate the parameters from the incomplete data set, and is a very simple and practical learning algorithm. This method can be widely applied to process missing data, truncated data and so-called noisy incomplete data. Each iteration of the EM algorithm consists of two steps: e, step E: solving the expectation; and M, solving the maximum.
Disclosure of Invention
The invention aims to provide an application method of a mixed expert system with better prediction performance on lung cancer types in lung adenocarcinoma classification.
In order to achieve the above object, the present invention provides a method for applying a hybrid expert system to lung adenocarcinoma classification, comprising the steps of:
step 1: sorting the original data;
step 2: screening characteristic data with classification significance by adopting LASSO algorithm;
step 3: dividing the characteristic data into a training set and a testing set according to the proportion of 7:3;
step 4: performing five-fold cross validation on an ME or HME system in a training set to obtain an optimal model of the ME or HME system;
step 5: and testing the optimal model of the ME or HME system in a test set.
Preferably, in step 1, the raw data are image features and clinical features of the patient.
Preferably, in step 2, the LASSO algorithm changes coefficients of some features which do not have significant effect on prediction to 0 by adding a regularization term of an L1 norm to the generalized linear model, so as to achieve the effect of feature screening.
Preferably, in step 4, the five-fold cross validation is to divide the training set into five parts randomly, train the ME or HME system with four parts of the training set by specifying different super parameters, and validate the ME or HME system under different super parameters with the remaining one part; and selecting the super-parameters with optimal performance through the verification set, thereby establishing an optimal ME or HME system model.
Preferably, in step 5, the optimal model is tested in the test set and its AUC value is calculated as the true predictive performance of the model.
Preferably, the algorithm flow of the ME and HME system is as follows:
step a: inputting classified significant characteristic data screened by LASSO algorithm;
step b: adjusting model structure parameters and model parameter initial values;
step c: the characteristic data is processed by the model structure parameters and model parameter initial values to generate current parameters;
step d: e, the current parameters enter an E step of an EM algorithm, a Q function is obtained through calculation, and a final objective optimization function value is obtained after the regular term parameters are added: j function value.
Step e: judging whether the J function value is a bracelet or not, if yes, outputting current model parameters, and if no, entering M steps of an EM algorithm;
step f: and M step of EM algorithm, namely partial differentiation is carried out on all the parameters respectively by the J function values, then the J function is updated to new parameters by using RMSProp algorithm, the step C is returned to carry out circulation until the step e is judged to be converged, and the step E is output.
Preferably, in step b, the model structure parameters include: the number of bottom layer models, RMSProp algorithm parameters, the number of batches of small batch random gradient descent and regularization coefficients;
the initial value of the model parameter is the initial value of the network parameter to be trained of the ME and HME systems, and a random value between 0 and 1 is taken according to the shape of the initial value.
Preferably, in step c, the current year parameter further includes a parameter updated by the RMSProp algorithm in step f.
Preferably, in step d, the Q function is a model parameter θ and a current iteration step model parameter θ (k) Is a function of (2);
the Q function for ME system is:wherein->Indicating the expected probability that the jth data will primarily use the jth underlying model.
For HME systems:wherein->Representing the expected probability that the (i, j) th underlying model is used primarily by the t-th data.
At the time of adding positiveThen the term parameters are followed to obtain the final objective optimization function: j (theta ) (k) )＝-Q(θ，θ (k) )+α[∑ j ||W j || 2 2 +∑ j (||W 1j || F 2 +||W 2j || 2 2 )]。
The Q function for ME system is:wherein->Indicating the expected probability that the jth data will primarily use the jth underlying model.
For HME systems:wherein->Representing the expected probability that the (i, j) th underlying model is used primarily by the t-th data.
After adding the regularization term parameters, a final objective optimization function is obtained: j (theta ) (k) )＝-Q(θ，θ (k) )+α[∑ j ||W j || 2 2 +∑ j (||W 1j || F 2 +||W 2j || 2 2 )]。
Compared with the prior art, the invention has the advantages that: compared with a single statistical model and a machine learning method, the application method of the mixed expert system in lung adenocarcinoma classification has better prediction performance on lung cancer types.
Drawings
FIG. 1 is a flow chart of a hybrid expert system in an embodiment of the invention;
FIG. 2 is a flow chart of a hierarchical hybrid expert system in accordance with an embodiment of the present invention;
FIG. 3 is a flowchart illustrating an algorithm for an ME and HME system in accordance with one embodiment of the invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be further described below.
In the present invention, for the convenience of understanding by those skilled in the art, as shown in fig. 1, the following explanation is made for the system of mixing specialists:
we first present some independent expert models, typically relatively simple and mature models, such as generalized linear models, low-level neural network models, etc. We use f j To represent these expert models, each of which can independently give a respective output for a given input x: u (u) j ＝f j (x)。
We call the network that regulates the weight of each expert model a gating network. In modeling, we assume that the gating network is broadly linear. Thus we define intermediate variables:wherein v is j Is a weight vector
The j-th output of the gating network is ζ j "softmax" function of:after the expert output and the gating network are obtained, the output of the final model is the weighted sum of the outputs of all the experts: u= Σ j g j u j 。
We can consider the hybrid expert system as a probabilistic generating model, i.e. the total probability of generating output y from input x is a mixture of probabilities from each component density generating y, where the mixing ratio is given in g in the gating network j Values. Let θ j For parameters in each expert model, v j For parameters in the gating network, the total probability is generated by: p (y|x, Θ) = Σ j g j (x，v j )P(y|x，θ j ). Wherein Θ includes expert model parameters θ j Gating network parameter v j 。
As shown in fig. 2, the following explanation is made for the system of layered hybrid experts:
the hybrid expert system is a special case of a hierarchical hybrid expert system layer number of 1. In particular, we take a two-tier hybrid expert system as an example, which is divided into a top-tier gating network and a lower-tier gating network. We use f ij Representing model functions in the expert network (i, j), each expert model can independently give a respective output for a given input x: u (u) ij ＝f ij (x)。
The gating network of the model is assumed to be also broadly linear. Then for the top-level gating network, ζ i And g i Is defined as:for low-level gating networks, ζ ij And g ij Is defined as: />Wherein v is i And v ij Are weight vectors of the input features. The output vector at each non-terminal node of the tree is a weighted output of the expert below the non-terminal node. That is, the output of the ith non-termination node of the second level of the tree is: u (u) i ＝∑ j g ij u ij And the output of the tree top layer is: u= Σ i g i u i 。
Similarly, consider the entire model as a probability generating model, let θ ij For parameters in each expert model, v i V is a parameter in the top-level gating network ij For parameters in the lower-layer gating network, the mixing ratio is given as g in the gating network i G ij The value, the total probability is generated by: p (y|x, Θ) = Σ i g i ∑ j g ij P(y|x，θ ij )。
Wherein Θ includes expert model parameters θ ij Gating network parameter v i V ij 。
In this embodiment, a method for applying a hybrid expert system to lung adenocarcinoma classification is provided, including the following steps:
step 1: raw data are arranged, wherein the raw data are image characteristics and clinical characteristics of a patient; a total of 13 are shown in the following table.
Step 2: screening characteristic data with classification significance by adopting LASSO algorithm;
since the input of the model is a numerical variable, all of the factorial variables are converted into numerical values (positive integers starting from 0). Then, in order to eliminate the influence of the difference of the value ranges of different variables, all variables are normalized to be between 0 and 1.
However, not all variables contribute to the final prediction, we use the LASSO algorithm to filter out features that are of categorical significance. The LASSO algorithm changes coefficients of some features which have no significant effect on prediction into 0 by adding a regularization term of L1 norm after the generalized linear model, so that the effect of feature screening is achieved.
The characteristics selected by the method comprise average diameter, average CT value, lung segment number, tumor lung interface and cavitation sign, and all have good prediction performance.
Step 3: dividing the characteristic data into a training set and a testing set according to the proportion of 7:3; if our model is trained on only all data, it will achieve good predictive performance on this dataset. But for new data externally it will lose its original performance, since it may only remember the original data, not actually have the ability to predict, which is also known as overfitting.
Therefore, we divide the total data into training and testing sets in a ratio of 7:3; after the model has been trained on the training set, its performance is evaluated on the test set.
Here, AUC (Area Under Curve) is used to evaluate the model, which is a performance index for measuring the quality of learners
Step 4: performing five-fold cross validation on the ME or HME system in the training set to obtain an optimal model of the ME or HME system; the five-fold cross verification is to divide the training set into five parts randomly, train the ME or HME system by four parts of the training set by designating different super parameters, and verify the ME or HME system under different super parameters by the rest one part; and selecting the super-parameters with optimal performance through the verification set, thereby establishing an optimal ME or HME system model.
Step 5: testing an optimal model of the ME or HME system in a test set; and testing the optimal model in the test set, and calculating the AUC value of the optimal model to serve as the real prediction performance of the model.
In this embodiment, as shown in fig. 3, the algorithm flow of the ME and HME system is as follows:
step a: inputting classified significant characteristic data screened by LASSO algorithm;
step b: adjusting model structure parameters and model parameter initial values;
step c: the characteristic data is processed by the model structure parameters and the model parameter initial values to generate current parameters;
step d: e, the current parameters enter an E step of an EM algorithm, a Q function is obtained through calculation, and a final target optimization function value is obtained after the regular term parameters are added: j function value.
Step e: judging whether the J function value is a bracelet or not, if yes, outputting current model parameters, and if no, entering an M step of an EM algorithm;
step f: and M step of EM algorithm, namely partial differentiation is carried out on all the parameters respectively by J function values, then the J function is updated to new parameters by RMSProp algorithm, the step C is returned to carry out circulation until the step e is judged to be converged and then the step e is output.
In this embodiment, in step b, the model structure parameters include: the number of bottom layer models, RMSProp algorithm parameters, the number of batches of small batch random gradient descent and regularization coefficients;
the initial value of the model parameter is the initial value of the network parameter which needs to be trained by the ME and HME systems, and the random value between 0 and 1 is taken according to the shape of the initial value.
In this embodiment, in step c, the current year parameter further includes the parameter updated by the RMSProp algorithm in step f.
In the present embodiment, in step d, the Q function is set to be related to the model parameter θ and the current iteration step model parameter θ (k) Is a function of (2);
the Q function for ME system is:wherein->Indicating the expected probability that the jth data will primarily use the jth underlying model.
For HME systems:wherein->Representing the expected probability that the (i, j) th underlying model is used primarily by the t-th data.
After adding the regularization term parameters, a final objective optimization function is obtained: j (theta ) (k) )＝-Q(θ，θ (k) )+α[∑ j ||W j || 2 2 +∑ j (||W 1j || F 2 +||W 2j || 2 2 )]。
The Q function for ME system is:wherein->Indicating the expected probability that the jth data will primarily use the jth underlying model.
For HME systems:wherein->Indicating that the (i, j) th bottom mode is mainly used for the (t) th dataExpected probability of type.
After adding the regularization term parameters, a final objective optimization function is obtained: j (theta ) (k) )＝-Q(θ，θ (k) )+α[∑ j ||W j || 2 2 +∑ j (||W 1j || F 2 +||W 2j || 2 2 )]。
Compared with a single statistical model and a machine learning method, the application method of the mixed expert system in lung adenocarcinoma classification has better prediction performance on the type of lung cancer; because the mixed expert system follows the standard flow of the film-taking of the image doctors, that is, a plurality of experts perform judgment at the same time, and finally, the opinions of all the experts are integrated, wherein the more experienced experts can obtain more weight.
The foregoing is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Any person skilled in the art will make any equivalent substitution or modification to the technical solution and technical content disclosed in the invention without departing from the scope of the technical solution of the invention, and the technical solution of the invention is not departing from the scope of the invention.
Claims (5)
1. A method of using a hybrid expert system in lung adenocarcinoma classification, comprising the steps of:
step 1: sorting the original data;
step 2: screening characteristic data with classification significance by adopting LASSO algorithm;
step 3: dividing the characteristic data into a training set and a testing set according to the proportion of 7:3;
step 4: performing five-fold cross validation on an ME or HME system in a training set to obtain an optimal model of the ME or HME system;
step 5: testing the optimal model of the ME or HME system in a test set;
the algorithm flow of the ME and HME systems is as follows:
step a: inputting classified significant characteristic data screened by LASSO algorithm;
step b: adjusting model structure parameters and model parameter initial values;
step c: the characteristic data is processed by the model structure parameters and model parameter initial values to generate current parameters;
step d: e, the current parameters enter an E step of an EM algorithm, a Q function is obtained through calculation, and a final objective optimization function value is obtained after the regular term parameters are added: j, a function value;
step e: judging whether the J function value is converged or not, if so, outputting current model parameters, and if not, entering an M step of an EM algorithm;
step f: m step of EM algorithm, calculate partial differentiation to all parameters of said J function value separately, then use RMSProp algorithm to update J function to the new parameter, return to step C and circulate, until step e judges to output after convergence;
in step b, the model structure parameters include: the number of bottom layer models, RMSProp algorithm parameters, the number of batches of small batch random gradient descent and regularization coefficients;
the initial value of the model parameter is an initial value of a network parameter to be trained of the ME and HME system, and a random value between 0 and 1 is taken according to the shape of the initial value;
in step c, the current parameters further include parameters updated by the RMSProp algorithm in step f;
in step d, the Q function is a model parameter θ and a current iteration step model parameter θ (k) Is a function of (2);
the Q function for ME system is:wherein->Representing expected probabilities that the jth data uses primarily the jth underlying model; g j Represents the gating weight, P (y (t) |x (t) ,θ j ) Represented by the t-th numberInput of data x (t) And the current parameter θ of the jth underlying model j Obtain the output y of the t-th data (t) Probability of (2);
for HME systems:wherein->Representing the expected probability that the (i, j) th underlying model is used primarily by the (t) th data; g ij Represents the gating weight, P (y), of the (i, j) th floor model (t) |x (t) ,θ ij ) Representing the input x of data by t (t) And the current parameter θ of the (i, j) th floor model ij Obtain the output y of the t-th data (t) Probability of (2);
after adding the regularization term parameters, a final objective optimization function is obtained: j (theta ) (k) )＝-Q(θ,θ (k) )+α[∑ j ||W j || 2 2 +∑ j (||W 1j || F 2 +||W 2j || 2 2 )]The method comprises the steps of carrying out a first treatment on the surface of the Alpha represents the weight of the regularized term parameter, W j Weight parameters representing a first layer expert model, W 1j And W is 2j And linear parameters representing a second layer generalized linear expert model.
2. The method of claim 1, wherein in step 1, the raw data is an image feature and a clinical feature of the patient.
3. The method according to claim 1, wherein in step 2, the LASSO algorithm changes coefficients of features that do not significantly contribute to the prediction to 0 by adding regularization terms of L1 norms to the generalized linear model, thereby achieving feature screening.
4. The method of claim 1, wherein in step 4, the five-fold cross-validation is to divide the training set into five parts randomly, and the four parts are used to train ME or HME systems by specifying different super parameters, and the remaining one part is used to validate ME or HME systems under different super parameters; and selecting the super-parameters with optimal performance through the verification set, thereby establishing an optimal ME or HME system model.
5. The method of using a hybrid expert system in lung adenocarcinoma classification as claimed in claim 1, wherein in step 5 the optimal model is tested in the test set and its AUC values are calculated as the true predictive performance of the model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910782827.5A CN110706804B (en) | 2019-08-23 | 2019-08-23 | Application method of mixed expert system in lung adenocarcinoma classification |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910782827.5A CN110706804B (en) | 2019-08-23 | 2019-08-23 | Application method of mixed expert system in lung adenocarcinoma classification |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110706804A CN110706804A (en) | 2020-01-17 |
| CN110706804B true CN110706804B (en) | 2024-02-02 |
Family
ID=69193833
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910782827.5A Active CN110706804B (en) | 2019-08-23 | 2019-08-23 | Application method of mixed expert system in lung adenocarcinoma classification |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110706804B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109528198A (en) * | 2019-01-16 | 2019-03-29 | 郑志明 | Brain response comparative approach based on low order Multivariate linear model |
| CN114120045B (en) * | 2022-01-25 | 2022-05-31 | 北京猫猫狗狗科技有限公司 | Target detection method and device based on multi-gate control hybrid expert model |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015066564A1 (en) * | 2013-10-31 | 2015-05-07 | Cancer Prevention And Cure, Ltd. | Methods of identification and diagnosis of lung diseases using classification systems and kits thereof |
| CN107220506A (en) * | 2017-06-05 | 2017-09-29 | 东华大学 | Breast cancer risk assessment analysis system based on depth convolutional neural networks |
| CN108231201A (en) * | 2018-01-25 | 2018-06-29 | 华中科技大学 | A kind of construction method, system and the application of disease data analyzing and processing model |
| CN108776962A (en) * | 2018-04-11 | 2018-11-09 | 浙江师范大学 | A method of the structure good pernicious prediction model of lung neoplasm |
| CN108921821A (en) * | 2018-06-01 | 2018-11-30 | 中国人民解放军战略支援部队信息工程大学 | Method of discrimination based on the LASSO mammary cancer armpit lymph gland transfering state returned |
| CN110111888A (en) * | 2019-05-16 | 2019-08-09 | 闻康集团股份有限公司 | A kind of XGBoost disease probability forecasting method, system and storage medium |
-
2019
- 2019-08-23 CN CN201910782827.5A patent/CN110706804B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015066564A1 (en) * | 2013-10-31 | 2015-05-07 | Cancer Prevention And Cure, Ltd. | Methods of identification and diagnosis of lung diseases using classification systems and kits thereof |
| CN107220506A (en) * | 2017-06-05 | 2017-09-29 | 东华大学 | Breast cancer risk assessment analysis system based on depth convolutional neural networks |
| CN108231201A (en) * | 2018-01-25 | 2018-06-29 | 华中科技大学 | A kind of construction method, system and the application of disease data analyzing and processing model |
| CN108776962A (en) * | 2018-04-11 | 2018-11-09 | 浙江师范大学 | A method of the structure good pernicious prediction model of lung neoplasm |
| CN108921821A (en) * | 2018-06-01 | 2018-11-30 | 中国人民解放军战略支援部队信息工程大学 | Method of discrimination based on the LASSO mammary cancer armpit lymph gland transfering state returned |
| CN110111888A (en) * | 2019-05-16 | 2019-08-09 | 闻康集团股份有限公司 | A kind of XGBoost disease probability forecasting method, system and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| A Mixture of Experts Network Structure for Breast Cancer Diagnosis;Elif Derya;《Journal of Medical Systems》;20051031;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110706804A (en) | 2020-01-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wyatt et al. | Anoddpm: Anomaly detection with denoising diffusion probabilistic models using simplex noise | |
| CN107316294B (en) | Lung nodule feature extraction method based on improved depth Boltzmann machine | |
| WO2018086433A1 (en) | Medical image segmenting method | |
| CN112101451B (en) | Breast cancer tissue pathological type classification method based on generation of antagonism network screening image block | |
| CN108257135A (en) | The assistant diagnosis system of medical image features is understood based on deep learning method | |
| Rao et al. | Convolutional neural networks for lung cancer screening in computed tomography (CT) scans | |
| CN110766051A (en) | Lung nodule morphological classification method based on neural network | |
| CN112270666A (en) | Non-small cell lung cancer pathological section identification method based on deep convolutional neural network | |
| CN108764358A (en) | A kind of Terahertz image-recognizing method, device, equipment and readable storage medium storing program for executing | |
| CN110706804B (en) | Application method of mixed expert system in lung adenocarcinoma classification | |
| CN108470337A (en) | A kind of sub- reality Lung neoplasm quantitative analysis method and system based on picture depth feature | |
| Popescu et al. | Retinal blood vessel segmentation using pix2pix gan | |
| CN112750137A (en) | Liver tumor segmentation method and system based on deep learning | |
| CN113989551A (en) | Alzheimer disease classification method based on improved ResNet network | |
| CN116386853A (en) | Intelligent medical-oriented deep separable convolution dual-aggregation federal learning method | |
| Yang et al. | Retinal vessel segmentation based on an improved deep forest | |
| Saikumar et al. | A Deep Convolutional Neural Network-Based Heart Diagnosis for Smart Healthcare Applications | |
| Hu et al. | A Hybrid Convolutional Neural Network Model Based on Different Evolution for Medical Image Classification. | |
| CN114372926A (en) | Traditional Chinese medicine tongue tenderness identification method based on image restoration and convolutional neural network | |
| CN117195027A (en) | Cluster weighted clustering integration method based on member selection | |
| CN112101438A (en) | Left and right eye classification method, device, server and storage medium | |
| CN111798463A (en) | Method for automatically segmenting multiple organs in head and neck CT image | |
| CN111353525A (en) | Modeling and missing value filling method for unbalanced incomplete data set | |
| CN114399501B (en) | Deep learning convolutional neural network-based method for automatically segmenting prostate whole gland | |
| de Oliveira Torres et al. | Texture analysis of lung nodules in computerized tomography images using functional diversity |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |